Spout

Turn your CSV data dictionary into a JSON repository. Collaborate with others to update the data dictionary in JSON format. Generate new Data Dictionary from the JSON repository. Test and validate your data dictionary using built-in tests, or add your own tests and validations.

Spout has been used extensively to curate and clean datasets available on the National Sleep Research Resource.

Add this line to your application's gems.rb:

gem "spout"

And then execute:

$ bundle

Or install it yourself as:

$ gem install spout

spout new my_data_dictionary

cd my_data_dictionary

spout import data_dictionary.csv

The CSV should contain at minimal the two column headers:

id: This column will give the variable its name, and also be used to name the file, i.e. <id>.json

folder: This can be blank, however it is used to place variables into a folder hiearchy. The folder column can contain forward slashes / to place a variable into a subfolder. An example may be, id: myvarid, folder: Demographics/Subfolder would create a file variables/Demographics/Subfolder/myvarid.json

Other columns that will be interpreted include:

display_name: The variable name as it is presented to the user. The display name should be fit on a single line.

description: A longer description of the variable.

type: Should be a valid variable type, i.e.:

• identifier
• choices
• integer
• numeric
• string
• text
• date
• time
• datetime
• file

domain: The name of the domain that is associated with the variable. Typically, only variable of type choices have domains. These domains then reside in domains folder.

units: A string of the associated that are appended to variable values, or added to coordinates in graphs representing the variable.

calculation: A calculation represented using algebraic expressions along with id of other variables.

labels: A series of different names for the variable that are semi-colon ; separated. These labels are commonly synonyms, or related terms used primarily for searching.

All other columns get grouped into a hash labeled other.

spout import data_dictionary_domains.csv --domains

The CSV should contain at minimal three column headers:

domain_id: The name of the associated domain for the choice/option.

value: The value of the choice/option.

display_name: The display name of the choice/option.

Other columns that are imported include:

description: A longer description of the choice/option.

folder: The name of the folder path where the domain resides.

spout import data_dictionary_domains.csv --forms

The CSV should contain at minimal three column headers:

folder: This can be blank, however it is used to place forms into a folder hiearchy. The folder column can contain forward slashes / to place a form into a subfolder. An example may be, id: family_history, folder: Demographics/BaselineVisit would create a file forms/Demographics/BaselineVisit/family_history.json

id: The reference name of the form.

display_name: The name of the form.

Other columns that are imported include:

code_book: The file name of the document or PDF, including the file extension.

If you created your data dictionary repository using spout new, you can go ahead and test using:

spout test

If not, you can add the following to your test directory to include all Spout tests, or just a subset of Spout tests.

test/dictionary_test.rb

require "spout/tests"

class DictionaryTest < Minitest::Test
  # This line includes all default Spout Dictionary tests
  include Spout::Tests
end

require "spout/tests"

class DictionaryTest < Minitest::Test
  # You can include only certain Spout tests by including them individually
  include Spout::Tests::JsonValidation
  include Spout::Tests::VariableTypeValidation
  include Spout::Tests::VariableNameUniqueness
  include Spout::Tests::DomainExistenceValidation
  include Spout::Tests::DomainFormat
  include Spout::Tests::DomainNameUniqueness
  include Spout::Tests::FormExistenceValidation
  include Spout::Tests::FormNameUniqueness
  include Spout::Tests::FormNameMatch
end

Then run either spout test or bundle exec rake to run your tests.

You can also use Spout iterators to create custom tests for variables, forms, and domains in your data dictionary.

Example Custom Test 1: Test that integer and numeric variables have a valid unit type

class DictionaryTest < Minitest::Test
  # This line includes all default Spout Dictionary tests.
  include Spout::Tests

  # This line provides access to @variables, @forms, and @domains iterators
  # that can be used to write custom tests.
  include Spout::Helpers::Iterators

  VALID_UNITS = ["minutes", "hours"]

  @variables.select { |v| %w(numeric integer).include?(v.type) }.each do |variable|
    define_method("test_units: #{variable.path}") do
      message = "\"#{variable.units}\"".red + " invalid units.\n" +
                "             Valid types: " +
                VALID_UNITS.sort_by(&:to_s).collect { |u| u.inspect.white }.join(", ")
      assert VALID_UNITS.include?(variable.units), message
    end
  end
end

Example Custom Test 2: Tests that variables have at least 2 or more labels.

class DictionaryTest < Minitest::Test
  # This line includes all default Spout Dictionary tests
  include Spout::Tests

  # This line provides access to @variables, @forms, and @domains
  # iterators that can be used to write custom tests
  include Spout::Helpers::Iterators

  @variables.select { |v| %w(numeric integer).include?(v.type) }.each do |variable|
    define_method("test_at_least_two_labels: #{variable.path}") do
      assert_operator 2, :<=, variable.labels.size
    end
  end
end

Spout lets you generate a nice visual coverage report that displays how well the data dictionary covers your dataset. Place your dataset csvs into ./csvs/<version>/ and then run the following Spout command:

spout coverage

This will generate an index.html file that can be opened and viewed in any browser.

Spout coverage validates that values stored in your dataset match up with variables and domains defined in your data dictionary.

Spout lets you generate detect outliers in your underlying datasets. Place your dataset csvs into ./csvs/<version>/ and then run the following Spout command:

spout outliers

This will generate an outliers.html file that can be opened and viewed in any browser.

Spout outliers computes the inner and outer fences to identify minor and major outliers in the dataset.

Provide an optional version parameter to name the folder the CSVs will be generated in, defaults to what is in VERSION file, or if that does not exist 1.0.0.

spout export

You can optionally provide a version string

spout export [1.0.0]

spout graphs

This command generates JSON charts and tables of each variable in a dataset

Requires a Spout YAML configuration file, .spout.yml, in the root of the data dictionary that defines the variables used to create the charts:

• visit: This variable is used to separate subject encounters in a histogram
• charts: Array of choices, numeric, or integer variables for charts

Example .spout.yml file:

visit: visitnumber
charts:
- chart: age
  title: Age
- chart: gender
  title: Gender
- chart: race
  title: Race

To only generate graphs for a few select variables, add the variable names after the spout graphs command.

For example, the command below will only generate graphs for the two variables ahi and bmi.

spout g ahi bmi

You can also specify a limit to the amount of rows to read in from the CSV files by specifying the -rows flag.

spout g --rows=10 ahi

This will generate a graph for ahi for the first 10 rows of each dataset CSV.

This will generate charts and tables for each variable in the dataset plotted against the variables listed under charts.

variables/Demographics/gender.json

{
  "id": "gender",
  "display_name": "Gender",
  "description": "Gender as reported by subject",
  "type": "choices",
  "domain": "gender12",
  "labels": [
    "gender"
  ],
  "commonly_used": true,
  "forms": [
    "intake_questionnaire"
  ]
}

domains/gender12.json

[
  {
    "value": "1",
    "display_name": "Male",
    "description": ""
  },
  {
    "value": "2",
    "display_name": "Female",
    "description": ""
  }
]

forms/Baseline Visit/intake_questionnaire.json

{
  "id": "intake_questionnaire",
  "display_name": "Intake Questionnaire at Baseline Visit",
  "code_book": "Baseline-Visit-Intake-Questionnaire.pdf"
}

spout deploy NAME

This command pushes a tagged version of the data dictionary to a webserver specified in the .spout.yml file.

webservers:
  - name: production
    url: https://sleepdata.org
  - name: staging
    url: https://staging.sleepdata.org

Shorthand

Deploy to Production

spout d p

Deploy to Staging

spout d s

The following steps are run:

• User Authorization
  • User authenticates via token, the user must be a dataset editor
• Version Check
  • "v#{VERSION}" matches HEAD git tag annotation
  • CHANGELOG.md top line should include version, ex: ## 0.1.0
  • Git Repo should have zero uncommitted changes
• Tests Pass
  • spout t passes for RC and FINAL versions
• Dataset Coverage Check
  • spout c passes for RC and FINAL versions
• Graph Generation
  • spout g is run
  • Graphs are pushed to server
• Dataset Uploads
  • Dataset CSV data dictionary is generated (variables, domains, forms)
  • Dataset and data dictionary CSVs uploaded to files section of dataset
• Documentation Uploads
  • README.md and KNOWNISSUES.md are uploaded
• Server-Side Updates
  • Server refreshes dataset folder to reflect new dataset and data dictionaries

You can check if a newer version of Spout is available by typing:

spout update